Initiation Factor 2 Crystal Structure Reveals a Different Domain Organization from Eukaryotic Initiation Factor 5B and Mechanism Among Translational Gtpases

Initiation factor 2 crystal structure reveals a different domain organization from eukaryotic initiation factor 5B and mechanism among translational GTPases

Daniel Eilera, Jinzhong Linb, Angelita Simonettic, Bruno P. Klaholzc, and Thomas A. Steitza,b,d,1

Departments of aChemistry and bMolecular Biochemistry and Biophysics, Yale University, New Haven, CT 06520; cCentre for Integrative Biology, Department of Integrated Structural Biology, Institute of Genetics and of Molecular and Cellular Biology, Centre National de la Recherche Scientiﬁque Unité Mixte de Recherche 7104, Institut National de la Santé de la Recherche Médicale U964, Université de Strasbourg, 67404 Illkirch, France; and dHoward Hughes Medical Institute, Chevy Chase, MD 20815

Edited by V. Ramakrishnan, Medical Research Council, Cambridge, United Kingdom, and approved August 9, 2013 (received for review May 22, 2013)

The initiation of protein synthesis uses initiation factor 2 (IF2) in which switch I, switch II and the P-loop are stimulated through prokaryotes and a related protein named eukaryotic initiation interaction with II and VI domains of the 23S rRNA in addition factor 5B (eIF5B) in eukaryotes. IF2 is a GTPase that positions the to L11 and L7/L12 of the large ribosomal subunit (1). GTP hy- initiator tRNA on the 30S ribosomal initiation complex and stim- drolysis by IF2 causes it to dissociate from the 70S IC and ulates its assembly to the 50S ribosomal subunit to make the 70S organizes the 70S ribosome for peptide elongation by a rotation ribosome. The 3.1-Å resolution X-ray crystal structures of the full- of the ribosomal subunits relative to each other (9–11). Cryo-EM length Thermus thermophilus apo IF2 and its complex with GDP studies of the 30S and 70S particles have shown that the final presented here exhibit two different conformations (all of its function of IF2 before it leaves the ribosome is to position the domains except C2 domain are visible). Unlike all other translational CCA end of the initiator tRNA near the peptidyl transferase GTPases, IF2 does not have an effecter domain that stably contacts center (6, 7, 9, 10). the switch II region of the GTPase domain. The domain organization Low-resolution cryo-electron microscopy (cryo-EM) recon- of IF2 is inconsistent with the “articulated lever” mechanism of com- structions have provided structural information about the inter- munication between the GTPase and initiator tRNA binding actions that govern the initiation process of 70S ribosome domains that has been proposed for eIF5B. Previous cryo-electron assembly in prokaryotes for protein synthesis. No structure from microscopy reconstructions, NMR experiments, and this structure full-length IF2 at atomic resolution has previously been de- show that IF2 transitions from being flexible in solution to an ex- termined. Until recently, the only structural information at tended conformation when interacting with ribosomal complexes. atomic resolution on IF2 has been NMR models of four separate domains of IF2 (N, G, C1, and C2) (12–15). The crystal structures of eukaryotic initiation factor 5B (eIF5B) he synthesis of proteins in prokaryotes is divided into three from Methanobacterium thermoautotrophicum have been used to Tdistinct processes: initiation, elongation, and termination. The construct a homology model of IF2 to interpret the electron initiation of translation in prokaryotes is directed by three initia- density from cryo-EM reconstructions of the 30S and 70S ICs with tion factors (IF1, IF2, and IF3) that govern the binding and po- IF2 (6, 7, 9, 10). Small angle X-ray scattering (SAXS) studies of E. sitioning of the mRNA, as well as the initiator tRNA, and the coli IF2 domains IV–VI (domains G, II, C1, and C2) indicate that joining of ribosomal subunits to form a 70S complex that is ready its domains are organized differently compared with eIF5B (16). for the elongation stage of protein synthesis. Allen et al. (10) found that the structure of the E. coli IF2 in their IF2 is a GTPase that functions to position the initiator tRNA cryo-EM reconstruction of a 70S IC differed significantly with that within the 30S ribosomal initiation complex (30S IC) and pro- of the crystal structure of M. thermoautotrophicum guanosine 5′- motes its joining with the 50S ribosomal subunit to form a 70S [β,γ-imido]triphosphateeIF5B. ribosome. IF2 is encoded by a single copy of the infB gene and is completely conserved in bacteria (1). The flexible structure of the N terminus has the largest sequence variability among dif- Significance ferent species (2). Variability also exists within a species; for instance, Escherichia coli IF2 has three isoforms, which vary in Initiation factor 2 (IF2) is a GTPase that functions within the 30S the length of their N-terminal domain due to three distinct start ribosomal initiation complex and promotes its joining with the sites for its translation initiation (1). The C-terminal part of IF2 50S ribosomal subunit to form a 70S ribosome. The role of IF2 (G, II, C1, and C2 domains) contains the highly conserved in translation initiation is not well understood. We present an GTPase domain (G domain) and C2 domain, which interact with atomic resolution crystal structure of the full-length IF2, and the initiator tRNA. we are able to explain why prokaryotes and eukaryotes have The C2 domain recognizes and protects the formylated Met of similar proteins with different mechanisms to guide ribosome the initiator tRNA from hydrolysis (3, 4). The formylation of assembly. We provide a structural explanation for why the Met results in a fivefold increase in the binding affinity of the tRNA and is made in a G-nucleotide–independent fashion (3, 5). mechanism of IF2 is unique among translational GTPases and This interaction permits IF2 to assist in positioning the initiator acts more as a novel conformational switch. tRNA within a 30S initiation complex and guide the formation of – Author contributions: D.E. and T.A.S. designed research; D.E. and J.L. performed research; a functional 30S IC on the establishment of the P-site codon D.E. and J.L. analyzed data; and D.E., J.L., A.S., B.P.K., and T.A.S. wrote the paper. anticodon interaction (2, 6, 7). A functional 30S IC is competent fl for the 50S ribosomal subunit to join, which is mediated by the The authors declare no con ict of interest. formation of the intersubunit salt bridges via an interaction be- This article is a PNAS Direct Submission. tween IF2 and L12 (8). Once the 50S ribosomal subunit joins, Freely available online through the PNAS open access option. IF2 comes into contact with the GTPase activation center, GTP Data deposition: The atomic coordinates and structure factors have been deposited in the is hydrolyzed by IF2, and IF2 dissociates from the ribosome (2, 9). Protein Data Bank, www.pdb.org (PDB ID code 4KJZ). IF2 is a GTPase homologous to other translational GTPases 1To whom correspondence should be addressed. E-mail: [email protected]. such as EF-Tu, EF-G, LepA, and RF3 (1). All translational This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. GTPases use a conserved mechanism for GTP hydrolysis in 1073/pnas.1309360110/-/DCSupplemental.

15662–15667 | PNAS | September 24, 2013 | vol. 110 | no. 39 www.pnas.org/cgi/doi/10.1073/pnas.1309360110 Downloaded by guest on October 1, 2021 Structural information is important for understanding, first Table 1. Data collection and refinement statistics (molecular how IF2 influences the position of the initiator tRNA in the 30S replacement) and 70S ICs, and second how IF2 facilitates joining of the ri- Data collection IF2 bosomal subunits to form a 70S ribosome ready for peptide

elongation. We address these aspects here and in a companion Space group P21 paper (17). Cell dimensions We determined the X-ray crystal structures of Thermus ther- a, b, c (Å) 119.83, 177.82, 61.79 mophilus IF2 from a full-length protein both with and without α, β, γ (°) 90.0, 88.87, 90.00 GDP bound at 3.09-Å resolution, which improves the model Resolution (Å) 50.0–3.09(3.21–3.09)* available for interpreting the cryo-EM reconstructions of the 30S Rsym 14.6 (>100) and 70S ICs. These two structures exhibit two different con- I/σI 10.2 (2) formations of IF2 (in apo and GDP-bound forms), and all of its Completeness (%) 100 (100) domains except C2 domain are visible. The initial molecular Redundancy 5.7 (5.8) replacement solution was determined by using a homology Refinement model based on an NMR model of the G domain from Geo- bacillus stearothermophilus Resolution (Å) 50–2.89 (PDB ID code 2LKD) and aligned fl with the G2 and G3 domains (IF2 G and II domains) of LepA No. re ections 57,096 from Aquifex aeolicus (PDB ID codes 2YWE, 2YWF, 2YWG, Rwork/Rfree 23.9/27.4 and 2YWH). While this work was being completed, a crystal No. atoms 13,732 structure of the first 363 residues of the T. thermophilus IF2 was Protein 13,659 determined and that model was then used as an improved search Ligand/ion 56 model for molecular replacement (18). Water 17 This structure of IF2 determined of residues 3–467 exhibits a B-factors different 3D organization compared with the homologous domains Protein 38.029 of eIF5B. We conclude that the organization of the IF2 domains Ligand/ion 50.524 indicates that the communication between the GTPase domain and Water 36.9 initiator tRNA binding domain cannot be explained by the “artic- R.m.s. deviations ulated lever” mechanism proposed previously for eIF5B. This Bond lengths (Å) 0.0165 difference in domain organization appears to reflect different Bond angles (°) 2.307 functions of the two proteins in translation initiation between prokaryotes and eukaryotes. Our crystal structure of isolated *Values in parentheses are for highest-resolution shell. IF2 shows that IF2 is unique among crystal structures of isolated translational GTPases because the effecter domains do not di- E. coli rectly contact or form a stable interaction with the switch II The cryo-EM reconstruction model of the 30S IC region of the GTPase domain. positions the N-terminal domain close to the A-site and near IF1, whereas this is not the case for T. thermophilus (7, 17). The Results N-terminal domain can be significantly longer in nonthermophilic Domain Organization. Most biochemical studies of IF2 have in- bacteria, and IF1 and IF2 interact with each other in the 30S IC, volved E. coli isoforms IF2-α (E. coli domains I–VI, full length) whereas the IF1 and IF2 interaction has been debated in or the more universally conserved IF2-β and -γ (E. coli domains T. thermophilus (22). The relative position of the G domain III–VI, lacking N1 and N2). Previous structural studies of IF2 may change from the 30S IC to the 70S IC conformation on have used a model centered on the crystal structure of eIF5B, 50S subunit joining. This change may cause the N-terminal which contains only the homologous G, II, C1, and C2 domains domain to lose some contacts with the ribosome and move into from bacteria. The crystal structure of IF2 from T. thermophilus the solvent contributing to IF2 adopting a “ready-to-leave” discussed here has a refined model including residues 3–467 of conformation seen elsewhere (9). 4 molecules of IF2 two GDP molecules and 17 water molecules (Table 1). Conserved GTPase Domains G and II. The fold of the GTPase domain of IF2 is similar to other G proteins that contain six- Overall Structure. The structure of IF2 bears a resemblance to stranded β-sheets enclosed by four α-helices (23). Two of the beads on a string, rather than exhibiting a chalice shape like copies in the asymmetric unit contain bound GDP but do not eIF5B. IF2 is likely to only adopt a stable and ordered confor- have switch I and II visible, which is consistent with the general mation when interacting with the ribosome, because it is known finding that switch II becomes ordered and adopts a helical to be a flexible protein. The structure presented here demon- conformation or is reordered when GTP is bound and interacts strates that long helices confer the element of flexibility. The specifically with the γ-phosphate, but is either disordered or has BIOCHEMISTRY overall dimensions of IF2 in the crystal are 65 × 88 Å for IF2 a different conformation when bound to GDP (23). GDP was not with GDP bound and 82 × 95 Å for apo IF2. For reasons dis- added, but was carried through from protein purification. Align- cussed later, the overall dimensions of IF2 in a functional active ment of the G domains of EF-Tu (PDB ID code 1EXM) on the state bound to the 30S ribosomal subunit are significantly larger. apo conformation of IF2 shows that the switch II structure in EF- Tu would be hindered by the packing of a C1 domain from N Domain. The N-terminal domain (N domain) is composed of a neighboring molecule of IF2 bound to GDP (Fig. 1) (24). a small, two-helix bundle that folds back onto a long extended Therefore, it appears that switch II is disordered and prevented helix (helix 3), which connects the N-terminal domain to the from forming a helical or other ordered structure because of the GTPase and G2 domains (G and II domains). The N domain, packing constrains on the C1 domain in the crystal. This packing including helix 3, functions in enhancing the interaction of IF2 also explains why our experiments to obtain IF2 crystals in with the 30S and 50S ribosomal subunits (17, 19, 20). Alignment complex with GTP by soaking or cocrystallization were un- of the G domains of IF2 (G and II domains) with that of EF-Tu successful. Based on the crystal structure alone, it is not apparent in complex with the 70S ribosome (PDB ID codes 2WRN and why GDP would bind to one conformation of IF2 in the asym- 2WRO) shows that helix 3 from the N-terminal domain of IF2 in metric unit and not to the other conformation. We see different the crystal is oriented into the solvent away from the 70S ribo- orientations of side chains when the polypeptide backbone of the some (21). The position of the N-terminal domain in the crystal GDP binding regions of the two conformations are super- is likely influenced by crystal packing. imposed, but at this resolution and accuracy, we cannot explain

Eiler et al. PNAS | September 24, 2013 | vol. 110 | no. 39 | 15663 Downloaded by guest on October 1, 2021 a conformation of IF2 where the C1 domain is closer to domains G and II. We expect that helix 8 and domain C1 are stabilized by the C2 domain interacting with the formylated initiator tRNA in the P-Loop ribosome. The orientation of the C2 domain, which is not visible in our crystal structure, indeed extends further from the C1 GTP domain in solution and on the 30S, as shown by SAXS and cryo- EM (17). The bend in helix 8 occurs at two different locations in the two IF2 conformations within the crystal. Switch II We believe that in solution, helix 8 of IF2 samples numerous Apo IF2 conformations and may not directly be affected by the absence or presence of GTP or GDP. The differences in C1 positions be- GTP·EF-Tu (1EXM) Steric Clash tween the GDP·IF2 and apo IF2 forms seen in the crystal and the IF2·GDP models determined from cryo-EM are shown in Fig. 2.

Fig. 1. G-nucleotide binding pocket of Apo IF2. G-nucleotide binding Comparison of IF2 with the eIF5B Structure. The eIF5B and IF2 pocket of apo IF2. The alignment of the G-domains of apo IF2 and EF-Tu·GTP models demonstrate conservation among domains as expected (1EXM) shows that ordering of the switch II region of IF2 on GTP binding (25). The G domain and the II, III, and IV domains of eIF5B would cause a steric clash with the neighboring molecule of IF2·GDP. correspond to the G, II, C1, and C2 domains of IF2, respectively. Minor differences between IF2 and eIF5B exist in the G domain, with two expansion segments being found in eIF5B (17). The how these changes cause the different conformations of the loop connecting strands 2 and 3 contains an 18 residue expansion domains. Although there is adequate space in the crystal for segment with three short helices, and a 32 residue expansion GDP to bind to the apo-form, soaking the crystals with up to 10 segment exists between β strands 5 and 6. mM GDP did not cause GDP to bind the apo conformation of Two major differences in the domain organization of IF2 IF2. Therefore, we conclude that because four copies of the and eIF5B are apparent. First, the most well known is the conformation of the GDP-bound IF2 cannot exist in the present presence of the variable N-terminal domain that is absent in crystal form, the structural changes that would happen in the M. thermoautotrophicum eIF5B. Other eukaryotes such as GDP binding pocket on GDP binding are precluded from oc- Saccharomyces cerevisiae and Homo sapiens have N-terminal curring within the conformation of apo IF2 in the crystal. domains that can be several hundred residues in length (26). The second, and more unexpected, is the effect of the length in the Helix 8 and C1 Domain. The IF2 crystal structure shows two distinct helix (24 residues longer than its equivalent in eIF5B) that orientations of the C1 domain among the four copies in the connects the II and C1 domains of IF2 compared with the helix asymmetric unit. The two copies of IF2 bound to GDP are that connects the II and III domains of eIF5B (Fig. 3). compact, and the C1 domain packs against residues 163–170 and A result of the second difference is that in the crystal struc- 200–209 of domain II. The two copies of apo IF2 are fanned out tures of these isolated factors, the G domain and switch II in the into an extended conformation and pack around residues 140, apo and GDP-bound forms of IF2 do not contact the C1 domain 170, 282, and helix 8 between domain II and the C1 domain of of IF2 as they do in the homologous domain III of eIF5B. In a symmetry mate. eIF5b, GTP hydrolysis on the G domain is proposed to be The difference between the two conformations exists in helix communicated to the C terminus using an articulated lever 8, which connects the II and C1 domains. The particular con- mechanism that involves domain III rotating with respect to the formation appears to be influenced by crystal packing but dem- G domains and changing the direction of the long helix (25). onstrates the flexibility of the helix 8 region. In the crystal, helix 8 Domain III rotates with respect to the G domains because switch in the structure of IF2 bound with GDP would sterically clash II becomes ordered or disordered when bound to GTP or with helix 8 of a second molecule of the IF2 bound with GDP if GDP, respectively. the helix 8 of either was further extended. This steric clash is Switch II in the G domain of IF2 does not contact the C1 prevented by a turn that occurs at residue 360. domain in the isolated crystal structure because the longer length A steric clash would also exist in the IF2 structure of the apo of helix 8 decouples the possibility of the G domain being able to communicate the state of G-nucleotide bound to the C2 domain form if helix 8 was continuous from residues 330–366, but the through switch II. This change in the length of helix 8 alters how bend in helix 8 is different. A clash with the second copy of the the C1 domain interacts with the G domain in the context of IF2 apo IF2 form is averted by a turn at residue 352. The region α crystal structure compared with the eIF5B crystal structure and reforms as an -helix at residue 358 and completes three helical raises questions about how GTP hydrolysis signal is communi- turns before the C1 domain, indicating a propensity for almost – α cated via switch II to the C1 domain in IF2. These dynamic the entire region of residues 330 366 of IF2 to form an -helix. regions are reorganized in the context of the ribosome as seen Helix 8 is a major structural and functional component, as this fl from the cryo-EM models in Fig. 2. These differences are sizable helix appears to be exible while also serving as the link between compared with the movement of 4.6 Å in the C terminus be- GTP hydrolysis and the C1 and C2 domains, the latter inter- tween Apo, GDP, and GTP forms of eIF5B (26). These differ- acting directly with the initiator tRNA. ences mean eIF5B and IF2 do not use the same articulated lever Discussion mechanism because of the different arrangement of their domains. In our structure, the helix connecting the C1 and C2 domains Flexibility of Helix 8 and Domain C1 in IF2 Is Supported by the Cryo- becomes disordered at residue 468, which is a conserved glycine EM Reconstructions. NMR experiments and cryo-EM recon- that lies in the middle of the equivalent helix (H12) in eIF5B. It structions support a model where IF2 is less flexible when is probable that this region becomes ordered when IF2 is bound interacting with ribosomal complexes (6, 7, 9, 10, 13). Cryo-EM to the ribosome and initiator tRNA is present. Furthermore, it is reconstructions of 30S and 70S ICs with IF2 bound show the of note that there are five fewer residues between the C1 and C2 overall structure of IF2 when it is bound to the ribosome with domains of IF2 compared with domains III and IV of eIF5B, some noticeable differences (6, 7, 9, 10). Simonetti et al. (6) based on the sequence length between the conserved domains. suggested a conformation in which the C1 domain is shifted The shorter length connecting the C1 and C2 domains may ex- toward the C2 domain away from the G and II domains while in plain the report of no trypsinization between the C1 and C2 the context of the 70S IC; Allen et al. (2005) suggested domains in IF2 (16).

15664 | www.pnas.org/cgi/doi/10.1073/pnas.1309360110 Eiler et al. Downloaded by guest on October 1, 2021 This work This work Accompanying 30S Cryo-EM 70S Cryo-EM 30S Cryo-EM T. thermophius T. thermophius Paper T. thermophius E. coli GDPNP-IF2 E. coli GDPNP-IF2 GDP-IF2 (2013) Apo IF2 (2013) 30S Cryo-EM GTP-IF2 (2008) (2005) (2011) T. thermophius GTP-IF2 (2013)

Fig. 2. Flexibility of C1 domain is apparent from crystal and cryo-EM models of IF2. C1 domains of IF2 are colored green, and the remainder is light gray. Models are from this work and refs. 6, 7, 10, and 17. The C1 domain is a ﬂexible element that is important to understanding the dynamic nature of IF2 and its unique GTPase activity.

These results mean that IF2 does not have the chalice struc- functional state that has lower affinity for the GDP-bound ture exhibited by eIF5B, and the mechanism of communication GTPase, which dissociates from the ribosome (30, 31). between GTP hydrolysis and initiator tRNA binding is different. IF2 shares features with EF-Tu, EF-G, and RF3. All of these translational GTPases require GTP hydrolysis to quickly dissociate Mechanism of IF2 Appears to be Unique Among the Translational from the ribosome. GTP analogs, such as Guanosine 5′-[β,γ-imido] GTPases of Prokaryotes. Most biochemical studies of IF2 have triphosphate and β,γ-Methyleneguanosine 5′-triphosphate, or muta- proposed that IF2 behaves as a classical GTPase. IF2 is active tions that eliminate GTPase activity will hinder the dissociation of when bound to GTP, carrying out its function as the steps of factors from the ribosome (30–33). However, IF2 differs from EF- hydrolysis transpire; it is inactive in its GDP-bound form and Tu, EF-G, and RF3 in a few respects. First, IF2 may not need exchanges GDP for GTP in solution without using an additional a nucleotide exchange factor (23, 30, 31, 34). Second, and more protein called a nucleotide exchange factor (27, 28). One well- unusual, IF2 has been shown to carry out its two functions, – characterized example is EF-Tu. EF-Tu forms a complex with positioning initiator tRNA and joining subunits, in a G-nucleotide GTP and aminoacyl tRNA and binds to the ribosome, and on independent method and even without a G-nucleotide (5, 28, 33, codon–anticodon recognition, GTP is hydrolyzed; EF-Tu then 35). IF2 is the only one of these GTPases used in translation that releases the aminoacyl tRNA for its accommodation in the does not have currently published atomic resolution data (crystal structure or NMR) showing direct contact between the functional A-site, and the EF-Tu complex with GDP dissociates from the domain and the switch II region of its GTPase domain (Fig. 4). ribosome. EF-Tu is an example of a classical GTPase that requires fi Whether interactions between IF2 and initiator tRNA take the protein EF-Ts, a nucleotide exchange factor, for ef cient place independently or only in the presence of the 30S ribosomal GDP/GTP exchange in solution. subunit or the 70S ribosome has been debated for decades (3, 6, Recently, a second hypothesis for the mechanism of IF2 has 34, 36–38). The stability of IF2 on the 30S subunit is strongest in been proposed in which IF2 acts as a conformational switch (29). the presence of both initiator tRNA and GTP, and the binding of A theory of conformational switching has also been proposed for GTP to IF2 is strongest in the presence of initiator tRNA and two other prokaryotic translational GTPases EF-G and RF3 30S ribosomal subunits (33, 35). It has also been shown that the fi (29). In the GDP-bound form, these proteins bind to speci c fastest transition from a 30S ribosomal initiation complex to fi functional states of the ribosome (30, 31). The speci c functional a 70S ribosome ready for peptide elongation is when IF2 bound state of the ribosome acts as a nucleotide exchange factor for with GTP is used (33, 35, 37). GDP to be replaced with GTP by a conformational change in Because the interaction of IF2 and initiator tRNA is strongest the domains of the GTPase (30, 31). According to the model, in the presence of the 30S ribosomal subunit, it is not GDP or changes occurring in the ribosome conformation because the GTP but the 30S ribosomal subunit that facilitates IF2 to interact new GTP-bound conformation of the GTPase favors GTP hy- with the initiator tRNA (6, 7). To fit IF2 into its ascribed elec- drolysis and only then is GTP hydrolyzed to produce a complex tron density within the cryo-EM reconstructions, IF2 needs to of GDP·EF-G or GDP·RF3 (30, 31). The ribosome is in a new make two contacts: the first contact is with the initiator tRNA,

B A eIF5b GDP·IF2 C Apo IF2 BIOCHEMISTRY II G-domain II II G-domain G-domain

Helix 8 352

Fig. 3. Comparison of helix 8 among eIF5b and IF2 Helix 8 352 crystal structures. Domain organization between III C1 360 360 eIF5b and IF2 conformations from this work. Ho- C1 mologous domains colored according to Roll-Mecak et al. (25). Domain C2 corresponding to VI from eIF5b was not visible in the electron density of IF2. N-terminal domain is located behind domain II and colored IV brown. (A)eIF5b(1G7S).(B) Compact IF2 model (GDP bound). (C) Extended IF2 model (apo). Residues 352 and 360 are indicated in the new crystal structures.

Eiler et al. PNAS | September 24, 2013 | vol. 110 | no. 39 | 15665 Downloaded by guest on October 1, 2021 and the second is between domain II and a region of the 16S ribosomal subunit is largely what facilities the interaction be- rRNA of the 30S ribosomal subunit. Both contacts require helix tween IF2 and the initiator tRNA. IF2 does not behave as a 8 to be in an elongated conformation. GTP-bound IF2 provides classical GTPase and acts more as a conformational switch that is stability based on kinetics experiments listed above; again, GTP heavily influenced by interactions with the initiator tRNA, the is not required as it is for EF-Tu, EF-G, and RF3 function. 30S ribosomal subunit, and the 70S ribosome. IF2 is not a classical GTPase and acts more as a conformational switch, although IF2 is not a conformational switch like Materials and Methods EF-G and RF3 have been proposed to be. IF2 functions better All chemicals except for those listed were bought from Sigma-Aldrich. Oli- with GTP but does not require it, and IF2 does not have an gonucleotides and the Quikchange mutagenesis kits were purchased from identified nucleotide exchange factor. One important structural Integrated DNA Technologies and Agilent, respectively. Crystallization question to be addressed is as follows: What is the significance of screens were bought from Hampton Research and Qiagen. DNA sequencing a nucleotide bound to IF2 in ribosomal complexes? To get was performed at the Keck Foundation Research Biotechnology Laboratory a complete picture of the mechanism, higher-resolution struc- (Yale University). Plasmid pET30b with an IF2 clone from Thermus thermophiles HB8 was tural information is needed on 30S and 70S ICs with IF2. obtained from A. E. Dahlberg (Brown University, Providence, RI). Quik- Conclusions change mutagenesis was performed to generate mutants according to the manufacturer’s guidelines. E. coli BL21(DE3) cells (Stratagene) transformed The structure of T. thermophilus IF2 that has been determined fl with the pET30b IF2 construct were grown at 37 °C in Luria Broth medium in here shows that much of the structural exibility in IF2 pre- the presence of 34 mg/L kanamycin to an absorbance at 600 nm of 0.5–0.7 viously described by other experiments is achieved through he- before induction with 1 mM isopropyl-β-D-thiogalactopyranoside. Cells were lices 3 and 8. The most important mechanistic consequences of grown for an additional 4 h before harvesting. The harvested cells were these two helices is helix 8, which does not allow for the GTP lysed, heat treated for 20 min at 65 °C, clarified by ultracentrifugation, and hydrolysis in domain II to be communicated to the C2 domain filtered. Ammonium sulfate was added to the sample (final concentration of that contacts the initiator tRNA. The longer length of this helix 1 M) and then loaded onto a phenyl Sepharose column (GE Heathcare). After is an important difference compared with eIF5B, where GTP a reverse gradient of ammonium sulfate (1–0 M), the IF2-containing frac- hydrolysis is communicated to the C terminus using an articu- tions were pooled, concentrated, and loaded onto a HiLoad 26/60 Superdex lated lever mechanism. This same mechanism cannot exist in 200 column (GE Heathcare) equilibrated in 30 mM Hepes KOH (pH 7.5), β both eIF5B and IF2 because of the differences in the lengths of 10 mM MgCl2,30mMNH4Cl, 1 mM EDTA NaOH, and 1 mM -mercaptoethanol. these helices. These differences in helix length lead to alterations The purified IF2 fractions from the gel-filtration column were again pooled and concentrated by ultrafiltration at a 10-kDa cutoff (Millipore) to 50–70 mg/mL as in domain contacts and therefore fundamentally change or fl eliminate the method of domain communication of GTP hy- determined by Abs280 (39). IF2 was ash-frozen in liquid nitrogen in small ali- quots and stored at −80 °C until further use in crystallization experiments. drolysis via switch II to the C2 domain via C1. Helix 8 is bent at two different places in the four molecules in Crystallization. WT IF2 crystallized as clusters of needles and rods and initially the asymmetric unit of the crystal structures determined here. To grew within 2 wk by vapor diffusion using a 1:1 ratio of 15 mg/mL IF2 to a well be consistent with the cryo-EM reconstructions of 30S and 70S solution containing 0.2 M calcium acetate, 0.1 M Na cacodylate, pH 6.5, and ICs, IF2 needs to have helix 8 act as a mobile structural element fi 18% (wt/vol) PEG 8000. The crystals of WT and the T17C mutant (originally and not as in the speci c static conformation seen in the crystal. designed for phasing attempts with heavy atom derivatives) were improved The determined crystal structure of IF2 provides structural by adjusting the well solution concentrations to 0.1 M calcium acetate, 0.04 M reasoning as to why IF2 interacts with initiator tRNA in- Na cacodylate, pH 5.4, 8% (wt/vol) PEG 8000, 10–30 mM glycl-glycine, and dependent of a G-nucleotide and is most stable in the presence 10–30 mM taurine, and adjusting the ratio to 4:1 of 25 mg/mL IF2 to well of the 30S ribosomal subunit. As previously suggested, the 30S solution. Rod-shaped crystals from clusters (WT and T17C) with dimensions

ABIF2·GDP EF-G·GDP (2EFG) C LepA·GDP (2YWH)

eIF5b·GDP (1G7S) DERF3·GDP (3VQT) F EF-Tu·GDP (1TUI)

Fig. 4. Comparison of switch II regions of translational GTPases. Homology domains colored by G2- GTPase (red), G3 domain (yellow), and GTP/GDP functional domains (green). The switch II region for each protein is circled in black, and the area of contact between switch II and its functional domains is indicated by a black arrow. IF2 is the only protein without switch II contacting a functional domain. (A) IF2 with GDP bound. (B) EF-G with GDP bound (PDB ID code 2EFG). (C) LepA with GDP bound (2YWH). (D) eIF5B with GDP bound (1G7S). (E) RF3 with GDP bound (3VQT). (F) EF-Tu with GDP bound (1TUI).

15666 | www.pnas.org/cgi/doi/10.1073/pnas.1309360110 Eiler et al. Downloaded by guest on October 1, 2021 of up to 800 × 200 × 25 μm were stabilized and cryoprotected by increasing In the model, 94.2% of all residues are in the favored, 5.5% are in the allowed, the concentration of PEG 8000 and ethylene glycol to a final concentration and 0.3% are in the outlier regions of the Ramachandran plot. All structural of 18% and 20% (wt/vol), respectively. Crystals were flash-frozen in liquid figures were prepared using PyMOL (www.pymol.org/). nitrogen. Full-length T. thermophilus IF2 crystallizes as a pseudomerohedral perfect

Diffraction data were collected at 100 K using X-rays at 0.97923 Å at the twin in the P21 space group. Conventional twinning tests yielded false- beamline 24ID-E at Advanced Photon Source in Argonne National Labora- negative results because of strong translational NCS and anisotropic dif- tory (Argonne, IL). The raw data were processed and scaled with the fraction. The space group was incorrectly assigned initially as orthorhombic HKL2000 program suite (40). General handling of scaled data was done us- due to the β angle being within an error of 90° and the crystal being per- ing Collaborative Computational Project programs (41) and Phenix (42). fectly twinned. Additional methods for the structure solution can be found in SI Materials and Methods. Structure Determination and Refinement. Full-length T. thermophilus IF2 crystallizes as a pseudomerohedral perfect twin in the P21 space group. The ACKNOWLEDGMENTS. We thank the staff at Brookhaven National Labora- asymmetric unit of the crystal contains four copies of IF2, in which the domains tory (beamline X25 and X29) and at the Advanced Photon Source in Argonne N, G, and II are related by a noncrystallographic symmetry (NCS) translation in National Laboratory (The Northeastern Collaborative Access Team 24ID-C the crystallographic asymmetric unit. The presence of strong translational NCS, and 24ID-E) for facilitating X-ray data collection. We thank former T.A.S. crystal twinning, and strong anisotropic diffraction made structure de- laboratory members G. Blaha and R. Stanley and current Yale Richards termination challenging. Crystallographic refinement statistics are pre- Center members Jimin Wang and Wuyi Meng for helpful discussion and advice. This work was supported by National Institutes of Health (NIH) sented in Table 1. Grant GM022778. D.E. was supported in part by NIH Predoctoral Program in Biophysics Grant NIH 5 T32 GM 8283-25. A.S. and B.P.K. acknowledge Structure Refinement. The model was built using COOT (43) with refinement in support from a European Research Council Starting Grant and French In- REFMAC5 (41). The structure was refined to Rfactor = 23.9% and Rfree = 27.4%. frastructure for Integrated Structural Biology Grant ANR-10-INSB-05-01.

1. Margus T, Remm M, Tenson T (2007) Phylogenetic distribution of translational 23. Sprang SR (1997) G protein mechanisms: Insights from structural analysis. Annu Rev GTPases in bacteria. BMC Genomics 8:15. Biochem 66:639–678. 2. Laursen BS, Sørensen HP, Mortensen KK, Sperling-Petersen HU (2005) Initiation of 24. Hilgenfeld R, Mesters JR, Hogg T (2000) The Ribosome: Structure, Function, Anti- protein synthesis in bacteria. Microbiol Mol Biol Rev 69(1):101–123. biotics, and Cellular Interactions (ASM Press, Washington, DC). 3. Petersen HU, Røll T, Grunberg-Manago M, Clark BF (1979) Specific interaction of 25. Roll-Mecak A, Shin BS, Dever TE, Burley SK (2001) Engaging the ribosome: universal initiation factor IF2 of E. coli with formylmethionyl-tRNA f Met. Biochem Biophys Res IFs of translation. Trends Biochem Sci 26(12):705–709. – Commun 91(3):1068 1074. 26. Roll-Mecak A, Cao C, Dever TE, Burley SK (2000) X-Ray structures of the universal 4. Spurio R, et al. (2000) The C-terminal subdomain (IF2 C-2) contains the entire fMet- translation initiation factor IF2/eIF5B: Conformational changes on GDP and GTP – tRNA binding site of initiation factor IF2. J Biol Chem 275(4):2447 2454. binding. Cell 103(5):781–792. 5. La Teana A, Pon CL, Gualerzi CO (1996) Late events in translation initiation. Adjust- 27. Bourne HR, Sanders DA, McCormick F (1991) The GTPase superfamily: Conserved ment of fMet-tRNA in the ribosomal P-site. J Mol Biol 256(4):667–675. structure and molecular mechanism. Nature 349(6305):117–127. 6. Simonetti A, et al. (2008) Structure of the 30S translation initiation complex. Nature 28. Tomsic J, et al. (2000) Late events of translation initiation in bacteria: A kinetic 455(7211):416–420. analysis. EMBO J 19(9):2127–2136. 7. Julián P, et al. (2011) The Cryo-EM structure of a complete 30S translation initiation 29. Hauryliuk V, Hansson S, Ehrenberg M (2008) Cofactor dependent conformational complex from Escherichia coli. PLoS Biol 9(7):e1001095. switching of GTPases. Biophys J 95(4):1704–1715. 8. Kay A, Sander G, Grunberg-Manago M (1973) Effect of ribosomal protein L12 upon 30. Zavialov AV, Buckingham RH, Ehrenberg M (2001) A posttermination ribosomal initiation factor IF-2 activities. Biochem Biophys Res Commun 51(4):979–986. 9. Myasnikov AG, et al. (2005) Conformational transition of initiation factor 2 from the complex is the guanine nucleotide exchange factor for peptide release factor RF3. – GTP- to GDP-bound state visualized on the ribosome. Nat Struct Mol Biol 12(12): Cell 107(1):115 124. 1145–1149. 31. Zavialov AV, Hauryliuk VV, Ehrenberg M (2005) Guanine-nucleotide exchange on ri- 10. Allen GS, Zavialov A, Gursky R, Ehrenberg M, Frank J (2005) The cryo-EM structure of bosome-bound elongation factor G initiates the translocation of tRNAs. J Biol 4(2):9. a translation initiation complex from Escherichia coli. Cell 121(5):703–712. 32. Luchin S, et al. (1999) In vitro study of two dominant inhibitory GTPase mutants of 11. Marshall RA, Dorywalska M, Puglisi JD (2008) Irreversible chemical steps control in- Escherichia coli translation initiation factor IF2. Direct evidence that GTP hydrolysis is tersubunit dynamics during translation. Proc Natl Acad Sci USA 105(40):15364–15369. necessary for factor recycling. J Biol Chem 274(10):6074–6079. 12. Laursen BS, Mortensen KK, Sperling-Petersen HU, Hoffman DW (2003) A conserved 33. Antoun A, Pavlov MY, Andersson K, Tenson T, Ehrenberg M (2003) The roles of ini- structural motif at the N terminus of bacterial translation initiation factor IF2. J Biol tiation factor 2 and guanosine triphosphate in initiation of protein synthesis. EMBO J Chem 278(18):16320–16328. 22(20):5593–5601. 13. Wienk H, et al. (2012) Structural dynamics of bacterial translation initiation factor IF2. 34. Milon P, et al. (2006) The nucleotide-binding site of bacterial translation initiation J Biol Chem 287(14):10922–10932. factor 2 (IF2) as a metabolic sensor. Proc Natl Acad Sci USA 103(38):13962–13967. 14. Wienk H, et al. (2005) Solution structure of the C1-subdomain of Bacillus stear- 35. Antoun A, Pavlov MY, Tenson T, Ehrenberg M (2004) Ribosome formation from – othermophilus translation initiation factor IF2. Protein Sci 14(9):2461 2468. subunits studied by stopped-flow and Rayleigh light scattering. Biol Proced Online 6: 15. Meunier S, et al. (2000) Structure of the fMet-tRNA(fMet)-binding domain of 35–54. – B. stearothermophilus initiation factor IF2. EMBO J 19(8):1918 1926. 36. Milon P, et al. (2010) The ribosome-bound initiation factor 2 recruits initiator tRNA to 16. Rasmussen LC, et al. (2008) Solution structure of C-terminal Escherichia coli translation the 30S initiation complex. EMBO Rep 11(4):312–316. initiation factor IF2 by small-angle X-ray scattering. Biochemistry 47(20):5590–5598. 37. Tsai A, et al. (2012) Heterogeneous pathways and timing of factor departure during 17. Simonetti A, et al. (2013) Involvement of protein IF2 N domain in ribosomal subunit translation initiation. Nature 487(7407):390–393. joining revealed from architecture and function of the full-length initiation factor. 38. Lockwood AH, Chakraborty PR, Maitra U (1971) A complex between initiation factor Proc Natl Acad Sci USA 110:15656–15661. IF2, guanosine triphosphate, and fMet-tRNA: An intermediate in initiation complex BIOCHEMISTRY 18. Simonetti AMS, et al. (2013) Structure of the protein core of translation initiation – factor IF2 in apo, GTP and GDP forms. Acta Crystallogr D Biol Crystallogr 69(6):925–33. formation. Proc Natl Acad Sci USA 68(12):3122 3126. fi 19. Moreno JM, Drskjøtersen L, Kristensen JE, Mortensen KK, Sperling-Petersen HU (1999) 39. Gill SC, von Hippel PH (1989) Calculation of protein extinction coef cients from amino – Characterization of the domains of E. coli initiation factor IF2 responsible for rec- acid sequence data. Anal Biochem 182(2):319 326. ognition of the ribosome. FEBS Lett 455(1–2):130–134. 40. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in os- – 20. Caserta E, et al. (2010) Ribosomal interaction of Bacillus stearothermophilus translation cillation mode. Methods Enzymol: Macromolec Crystallogr A 276:307 326. initiation factor IF2: Characterization of the active sites. J Mol Biol 396(1):118–129. 41. Winn MD, et al. (2011) Overview of the CCP4 suite and current developments. Acta 21. Schmeing TM, et al. (2009) The crystal structure of the ribosome bound to EF-Tu and Crystallogr D Biol Crystallogr 67(Pt 4):235–242. aminoacyl-tRNA. Science 326(5953):688–694. 42. Adams PD, et al. (2010) PHENIX: A comprehensive Python-based system for macro- 22. Kapralou S, et al. (2008) Translation initiation factor IF1 of Bacillus stear- molecular structure solution. Acta Crystallogr D Biol Crystallogr 66(Pt 2):213–221. othermophilus and Thermus thermophilus substitute for Escherichia coli IF1 in vivo 43. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot. and in vitro without a direct IF1-IF2 interaction. Mol Microbiol 70(6):1368–1377. Acta Crystallogr D Biol Crystallogr 66(Pt 4):486–501.

Eiler et al. PNAS | September 24, 2013 | vol. 110 | no. 39 | 15667 Downloaded by guest on October 1, 2021